Skip to content

Releases: boshu2/agentops

v3.0.1

25 May 13:02
a173082

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v3.0.1 is the recommended cut of the 3.0 major — the largest release in the project's history (~320 commits and a net rewrite, ~138K insertions / ~133K deletions since v2.41). 3.0 is the hookless, in-session-only rearchitecture: AgentOps is what runs inside a session — skills, the ao CLI, the RPI / evolve / crank / swarm loops, and the .agents/ context-compiler — and out-of-session orchestration (when work runs, who supervises it, how agents coordinate) is delegated to a substrate you choose: Gas City (a reference City ships here) or Mount Olympus.

The release removes hooks, the daemon, the scheduler, and the factory command (~117K LOC); consolidates the skill catalog 81→75 and gives every loop skill executable BDD acceptance; lands real hexagonal port adapters; ships an ao doctor repair engine and a much larger ao goals / ao evolve surface; and reconciles all doctrine to a single in-session north star (docs/3.0.md). 3.0.1 itself is the release-engineering patch that makes the plugin updatable and greens the nightly validation lane.

Upgrade Notes

  • New installs should use v3.0.1 (the clean cut). Existing v3.0.0 installs: run claude plugin update agentops@agentops-marketplace; if a plain update still reports "already latest", uninstall --keep-data then install once.
  • If you ran AgentOps in a session, nothing you rely on changed — ao rpi, ao evolve, crank, swarm, and the .agents/ corpus are unchanged.
  • If you ran the daemon, scheduler, or hooks, those are removed; see Breaking Changes below.
  • See docs/MIGRATION-3.0.md for the full breaking-change migration.

Breaking Changes

  • Hooks are removed (soc-57b7f). The default install registers no hooks; ao hooks install, --with-hooks, and the embedded hook surface are gone. What a hook enforced locally, a CI job in .github/workflows/validate.yml now enforces; author bounded gates with the hooks-authoring skill.
  • The daemon is removed (soc-2rtm0). ao daemon / agentopsd and internal/daemon no longer exist. For always-on work, run the reference Gas City (city.toml + packs/agentops) or Mount Olympus.
  • Scheduling is removed (soc-2rtm0). ao schedule, ao plans, ao watch, and ao overnight (commands + engines) are gone; Gas City Orders own out-of-session scheduling.
  • The factory command is removed (soc-2rtm0). ao factory and its contract corpus are gone; the factory is the loop (crank / swarm in-session, or a mayor-driven City).
  • runtime=gc is removed (soc-2rtm0). ao rpi keeps its auto / direct / stream / tmux backends; Gas City now dispatches whole ao rpi loops as one unit.

At a Glance

Product Area Added Changed Refactored Fixed Deprecated/Removed
Install, Upgrade, and Distribution 0 1 0 1 0
CLI and Operator Commands 6 1 0 2 4
Daemon, Scheduling, and Factory 0 0 0 0 4
Skills and Workflows 3 2 0 0 0
Codex and Runtime Integrations 0 1 0 0 0
Hooks and Lifecycle 0 0 0 0 1
Knowledge Flywheel, Search, and Memory 0 0 3 0 0
Eval, Validation, and Release Gates 5 2 0 2 0
Docs and Onboarding 4 2 0 0 0
Security, Privacy, and Supply Chain 1 1 0 0 0
Contributor/Internal Refactors 2 0 2 0 0

Product Areas

Install, Upgrade, and Distribution

  • Changed: version set across .claude-plugin/plugin.json, .claude-plugin/marketplace.json, and the City pack's pinned AO_VERSION — 3.0.0, then bumped to 3.0.1.
  • Fixed: claude plugin update stayed stale for early 3.0.0 installs because the version label never advanced past 3.0.0; the 3.0.1 bump gives the marketplace a new version to serve.

CLI and Operator Commands

  • Added: ao doctor — a shared diagnose-and-repair engine with subsystem detectors/fixers (cliconfig, daemon, bridges), a single --json Report, and offline-safe --fix / --dry-run exit codes.
  • Added: ao goals scenarios (--create / --lint / --list) for executable-spec links, plus ao goals steer --json now persists the mutation and steer add preserves non-directive content.
  • Added: ao validate --gate — an exit-code verdict (PASS/WARN/FAIL → process exit) for Gas City retry loops and CI.
  • Added: ao evolve config --show + a per-repo preferences.yaml schema/loader for the loop.
  • Added: ao loop surfaces for the HypothesisLedger and ConvergenceCheck ports.
  • Added: top-level capabilities + robot-docs agent surfaces, JSON subcommand listings, and --json on plans list / search / diff.
  • Changed: clearer error messages that name the corrective command (autodev PROGRAM.md, constraint/claim lookups).
  • Fixed: goals trace --orphans parser rejects bead prose and matches real orphans.
  • Fixed: a filesystem PacketRepository adapter behind the hexagonal packet seam.
  • Removed: ao daemon, ao schedule / plans / watch / overnight, ao factory, and the runtime=gc mode (see Breaking Changes).

Daemon, Scheduling, and Factory

  • Removed: internal/daemon + the agentopsd binary, the scheduler/overnight engines, the factory command + contract corpus, and the CLI gc-bridge glue (soc-2rtm0, ~117K LOC). AgentOps ships no out-of-session runtime; Gas City or Olympus owns that surface.

Skills and Workflows

  • Added: the AgentOps reference Gas City — city.toml + the agentops pack (mayor + refinery agents, skills overlay, Orders) + the using-gc agent guide and a complete tool-dependency declaration.
  • Added: canonical .feature acceptance across ~25 skills (research/plan/validation/crank/swarm/rpi/handoff/readme/flywheel/inject and more) plus SELF-TEST.md pilots.
  • Added: the idea-wizard generate-and-winnow flow folded into /brainstorm and the /discovery phase.
  • Changed: skill consolidation 81→75 — merged 4 skills into mode flags, cut 2, resolved 6 keep-flags.
  • Changed: /evolve hardening — --mode=burst|loop with cron self-adjust and typed blocked events, orchestrator-drive-to-completion (ADR-0008), never-sticky-dormancy while beads exist, auto-expiring KILL/STOP switches.

Codex and Runtime Integrations

  • Changed: the Codex skill twins (skills-codex/) and .codex-plugin manifest stayed in lockstep with the rip + consolidation, with hashes regenerated and parity audited across every skill change.

Hooks and Lifecycle

  • Removed: all 53 runtime hooks. The lifecycle is guided by skills and enforced by CI gates (see Breaking Changes).

Knowledge Flywheel, Search, and Memory

  • Refactored: a real corpus_fs CorpusReader/Writer adapter wired to its core consumer.
  • Refactored: a real bd-backed TrackerPort adapter wired to its core consumer.
  • Refactored: a real git-backed WorkspacePort adapter wired to the core worktree lifecycle.

Eval, Validation, and Release Gates

  • Added: a BDD acceptance layer — a scenario→test linkage runner plus a CI gate (soc-63xfx), with CLI-command and daemon-lifecycle Gherkin linked to tests.
  • Added: the SKU capability catalog (registry.json schema v2 — the generated skill×command×BC join) with a drift gate.
  • Added: a context-map drift gate and a validate-skill-body-refs gate for inline ao references in prose.
  • Added: validate-release-notes.sh as a hard release gate (curated notes must exist and conform to the standard).
  • Added: ci-local-release.sh --quick — the fast code-correctness subset for iterative pre-tag work.
  • Changed: executable-spec-link-integrity promoted warn→blocking; the scenario→test link gate is now required.
  • Changed: CI path-filters so validate-* and security-* jobs skip when irrelevant.
  • Fixed: Nightly Static Validation — removed the deleted validate-hooks-doc-parity.sh step and the hook-era PRODUCT.md skill-count patterns.
  • Fixed: pre-existing shellcheck warnings and post-merge CI failures cleared across the DDD/hex arc.

Docs and Onboarding

  • Added: docs/3.0.md — the canonical 3.0 north star, woven into the sources of truth.
  • Added: ADR-0009 (daemon deletion / in-session-only), ADR-0008 (orchestrator drive-to-completion), and the canonical-loop model.
  • Added: docs/MIGRATION-3.0.md — the breaking-change migration guide.
  • Added: the tiered AGENTS.md split, the sovereignty-proof page, and the release-notes standard (with a major-release variant).
  • Changed: README cut of AI-slop and restructure; PRODUCT and peripheral docs reconciled to the in-session core.
  • Changed: stale daemon / schedule / hook prose swept across narrative docs, with the out-of-session autonomous-dispatch gap explicitly labeled (soc-5jwah).

Security, Privacy, and Supply Chain

  • Added: executable acceptance for the repository security scans (security hexagon).
  • Changed: the pre-publish release evidence lane (CycloneDX SBOM + provenance attestation + the security-gate summary) now gates GoReleaser, blocking only on real security findings.

Contributor/Internal Refactors

  • Added: scripts/ship.sh (one-knob ship loop), the add-validate-job CI scaffolder, cron-tune-cadence, export-session-summary, and doctor-evolve preflight tooling.
  • Added: the domain ubiquitous-language register (architecture terms + behavior-shaping ABC vocabulary).
  • Refactored: the DDD / hexagonal architecture rescope across the bounded contexts.
  • Refactored: generated-artifact regeneration (registry, context-map, codex hashes) wired behind drift gates.

Known Issues

  • Gas City order-leve...
Read more

v3.0.0

25 May 04:18
caa635a

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v3.0.0 is the hookless, in-session-only major. The thesis sharpened: AgentOps is what runs inside a session — skills, the ao CLI, the RPI / evolve / crank / swarm loops, and the .agents/ context-compiler. Everything that tried to be an always-on orchestration runtime is gone, and out-of-session orchestration (when work runs, who supervises it, how agents coordinate) is delegated to a substrate you choose: Gas City (a reference City ships here) or Mount Olympus. Roughly 117K lines were removed across a five-wave rip, the default install now registers zero hooks, and CI is the authoritative gate. docs/3.0.md is the north star.

Upgrade Notes

  • If you ran AgentOps in a session, nothing you rely on changed — ao rpi, ao evolve, crank, swarm, and the .agents/ corpus are unchanged.
  • If you ran the daemon, scheduler, or hooks, those are removed; see Breaking Changes below and the migration guide.
  • See docs/MIGRATION-3.0.md for the full breaking-change migration.

Breaking Changes

  • Hooks are removed (soc-57b7f). The default install registers no hooks; ao hooks install, --with-hooks, and the embedded hook surface are gone. What a hook enforced locally, a CI job in .github/workflows/validate.yml now enforces; author bounded gates with the hooks-authoring skill.
  • The daemon is removed (soc-2rtm0). ao daemon / agentopsd no longer exist. For always-on work, run the reference Gas City (city.toml + packs/agentops) or Mount Olympus.
  • Scheduling is removed (soc-2rtm0). ao schedule, ao plans, ao watch, and ao overnight are gone; Gas City Orders own out-of-session scheduling.
  • The factory command is removed (soc-2rtm0). ao factory and its contract corpus are gone; the factory is the loop (crank / swarm in-session, or a mayor-driven City).
  • runtime=gc is removed (soc-2rtm0). ao rpi keeps its auto / direct / stream / tmux backends; Gas City now dispatches whole ao rpi loops as one unit.

At a Glance

Product Area Added Changed Refactored Fixed Deprecated/Removed
Install, Upgrade, and Distribution 0 1 0 0 0
CLI and Operator Commands 1 0 0 0 3
Daemon, Scheduling, and Factory 0 0 0 0 4
Skills and Workflows 1 1 0 0 0
Hooks and Lifecycle 0 0 0 0 1
Knowledge Flywheel, Search, and Memory 0 0 3 0 0
Eval, Validation, and Release Gates 2 0 0 0 0
Docs and Onboarding 3 1 0 0 0

Product Areas

Install, Upgrade, and Distribution

  • Changed: version synced to 3.0.0 across .claude-plugin/plugin.json, .claude-plugin/marketplace.json, and the City pack's pinned AO_VERSION.

CLI and Operator Commands

  • Added: ao validate --gate collapses PASS/WARN/FAIL to a process exit code — the retry hook for Gas City check and CI, no network or LLM.
  • Removed: ao daemon, ao schedule / plans / watch / overnight, and ao factory (see Breaking Changes).
  • Removed: the runtime=gc phased-engine mode.

Daemon, Scheduling, and Factory

  • Removed: internal/daemon + the agentopsd binary, the scheduler lane, the factory command + contract corpus, and the CLI gc-bridge glue (soc-2rtm0, ~117K LOC). AgentOps ships no out-of-session runtime; Gas City or Olympus owns that surface.

Skills and Workflows

  • Added: the AgentOps reference Gas City — a turnkey City (city.toml + packs/agentops + skills overlay) demonstrating out-of-session orchestration on AgentOps skills.
  • Changed: skill consolidation merged overlapping skills into mode flags, taking the catalog to 75 skills.

Hooks and Lifecycle

  • Removed: all 53 runtime hooks. The lifecycle is guided by skills and enforced by CI gates (see Breaking Changes).

Knowledge Flywheel, Search, and Memory

  • Refactored: real hexagonal port adapters wired to their core consumers — a filesystem CorpusReader/Writer, a bd-backed TrackerPort, and a git-backed WorkspacePort.

Eval, Validation, and Release Gates

  • Added: a BDD acceptance layer — scenario→test linkage runner plus a CI gate (soc-63xfx), with canonical .feature acceptance across skills.
  • Added: ao validate --gate as a deterministic exit-code verdict for CI and Gas City retry loops.

Docs and Onboarding

  • Added: the 3.0 doctrine — docs/3.0.md north star, docs/adr/ADR-0009-daemon-deletion-in-session-only.md, and the canonical-loop model.
  • Added: docs/MIGRATION-3.0.md, the breaking-change migration guide.
  • Changed: README and PRODUCT reconciled to the in-session core, with the out-of-session autonomous-dispatch gap explicitly labeled (soc-5jwah).

Known Issues

  • Gas City order-level autonomous dispatch is still maturing upstream (soc-5jwah); today the reference City dispatches via a long-lived mayor agent (bd readygc sling) plus cron exec orders for maintenance.

Full changelog


Full changelog

AgentOps 3.0 is the hookless-first major. The headline: AgentOps is what runs in-session — skills + the ao CLI + the RPI/evolve/crank/swarm loops + the context-compiler. Out-of-session orchestration (scheduling, daemons, autonomous dispatch) is delegated to Gas City (a reference City ships in this release) or Olympus. The default install registers zero hooks; the lifecycle is driven by skills and CI gates. See docs/3.0.md for the north star.

⚠️ Breaking Changes

  • Hookless by default (soc-57b7f) — the default install registers no hooks. All 53 runtime hooks were audited and removed; capabilities that lived in hooks (standards injection, commit-review gating, noise injectors) are now CLI ports + CI gates. Opt back in with --with-hooks (AGENTOPS_INSTALL_HOOKS=1) or ao hooks install --force. Migration: if you relied on hook-driven behavior, switch to the CI gate or the equivalent ao subcommand.
  • Daemon carved out (soc-2rtm0, daemon-carve wave 5) — internal/daemon + the agentopsd binary were deleted. AgentOps no longer ships an out-of-session daemon; Gas City is the orchestration substrate. Builder cores that previously ran as daemon jobs (wiki build, llmwiki loop) keep their in-process / GC-callable core but lose the daemon-job-executor wrapper.
  • Scheduling/plans/watch/overnight commands removed (soc-2rtm0, kill-schedule-overnight) — ao schedule, ao plans, ao watch, and the overnight engines were deleted. Gas City owns scheduling. The phased RPI engine keeps its non-gc backends (auto/direct/stream/tmux); runtime=gc is no longer a valid mode.
  • factory command + contract corpus retired (soc-2rtm0, cascade-rip) — the ao factory command and its contract corpus were removed.
  • Gas City (gc) bridge severed (soc-2rtm0, wave 2) — the CLI gc-bridge glue (gc_bridge.go, gc_events.go, rpi_phased_gc.go) was deleted; the daemon-side GasCity client is a separate, retained surface.

Added

  • AgentOps reference Gas City — a turnkey City (city.toml + agentops pack + overlay) that demonstrates out-of-session orchestration on top of AgentOps skills.
  • ao validate --gate — exit-code verdict for Gas City retry loops and CI; PASS/WARN/FAIL collapses to a process exit code.
  • Real hexagonal port adapterscorpus_fs CorpusReader/Writer, a bd-backed TrackerPort, and a git-backed WorkspacePort, each wired to its core consumer. Port-realness audit in docs/architecture/.
  • BDD acceptance layer — scenario→test linkage runner + CI gate (soc-63xfx); canonical .feature acceptance added across skills (research/plan/validation/trace/handoff/readme/autodev/flywheel/inject and more) and daemon-lifecycle Gherkin linked to tests.
  • scripts/ship.sh (#346, soc-33uy) — single-command wrapper that auto-detects inventory-touching diffs and routes through the full pre-push gate (no --fast skip on skill/contract changes), preemptively running the regen sweep (sync-skill-counts, codex-hashes, domain-map, context-map, registry, sync-hooks). Mechanical fix for ship-loop anti-pattern #1.

Changed

  • Honest 3.0 doctrinedocs/3.0.md is the canonical north star: AgentOps is in-session; Gas City and Olympus own out-of-session. Doctrine, PRODUCT, and README were reconciled to the in-session core (skills + rpi + evolve + crank + swarm + context-compiler), with the out-of-session autonomous-dispatch gap explicitly labeled (soc-5jwah).
  • Version sync to 3.0.0.claude-plugin/plugin.json, .claude-plugin/marketplace.json (metadata + plugins[0]), and the City pack's pinned AO_VERSION (packs/agentops/assets/scripts/install-ao.sh) all bump to 3.0.0.
  • Coherent-arc PR rule (#348, soc-1lp1) — replaces "one scenario per PR" default in CLAUDE.md + AGENTS.md. The unit of a PR is one closable bead (or small-epic slice) with a single rollback semantic. Small epics (≤5 child beads, same surface) ship as one PR with N commits; large epics ship as N PRs sliced by scenario or wave. Updates ship-loop skill (Claude + Codex twins). Derived from the 2026-05-19 8-PR merge-arc burn-through.
  • skill-auditor rubric scoring (soc-ads5v) — adds a Pass 3 rubric-scoring stage.
  • Executable-spec-link-integrity promoted warn→blocking (soc-x7y9f) — the scenario→test link gate is now a required CI check.

Fixed

  • Daemon robustness hardening — bounded idempotency key + request decode guards (soc-scg3h), ...
Read more

v2.41.1

15 May 19:59

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v2.41.1 is a CI-hygiene patch for v2.41.0. The v2.41.0 tag was cut on a commit where three Validate checks were already failing — masked because the path-filtered CI marks untriggered jobs skipped rather than failed, so latent red on main was invisible until a later commit touched the triggering paths.

This patch makes main genuinely green: regenerated registry.json, a real markdown link for the extracted references/flags.md, a refreshed cli-command-surface-matrix eval baseline, and 7 Codex-notation fixes in skills-codex/. It also completes the --oscillation-sweep flag removal that v2.41.0 left half-done (Go callers were audited; shell/YAML/doc callsites were not).

No product behavior changes — this release is entirely build/CI/release-tooling fixes.

Upgrade Notes

  • No CLI or behavioral changes from v2.41.0. If you are already on v2.41.0, this patch only matters if you build from source or run the CI gates.
  • The v2.41.0 GitHub Release remains as published; v2.41.1 is the clean, CI-green release. New installs should use v2.41.1.

Fixed

  • Release CI shipped red in v2.41.0. Three Validate checks were failing on the tagged commit:
    • registry-checkregistry.json was stale; the v2.41 BC arc added 10 ao subcommands without regenerating it.
    • skill-integrityskills/validation/SKILL.md referenced the extracted references/flags.md with a backtick code-span instead of a markdown link; heal.sh --strict requires a real link.
    • contract-canaries — the cli-command-surface-matrix eval baseline was stale (top=64 sub=159 all=223, now top=70 sub=173 all=243); and the codex-native install canary flagged 7 /skillname slash-command references in skills-codex/{evolve,validation}/SKILL.md (Codex uses $skillname notation).
  • --oscillation-sweep removal completed. soc-1q1x deleted the ao defrag --oscillation-sweep flag in v2.41.0 but the audit only covered Go callers. Seven shell/YAML/doc callsites still referenced the dead flag, breaking the validate-flywheel-proof CI gate. Cleared from .github/workflows/nightly.yml, scripts/nightly-dream-cycle.sh, scripts/check-compile-oscillation.sh, docs/contracts/dream-run-contract.md, and the compile skill phase docs.
  • extract-release-notes.sh double-blank line. The awk extractor captured the blank line after the ## [VERSION] heading; wrapped in the <details> boilerplate it produced a double blank between <summary> and the first subsection. The script now strips leading/trailing blank lines from the changelog section.

See Also

  • Full per-commit detail: CHANGELOG.md (entry [2.41.1])
  • Release audit trail: docs/releases/2026-05-15-v2.41.1-audit.md
  • v2.41.0 notes: docs/releases/2026-05-14-v2.41.0-notes.md
  • Follow-up epic to prevent recurrence: bd soc-rw9y — "Make releases reliable"

Full changelog

Fixed

  • Release CI shipped red in v2.41.0 — the v2.41.0 tag was cut on a commit where three Validate checks were already failing; they were masked because the path-filtered CI marks untriggered jobs skipped (not failed). This patch clears all three: registry.json regenerated after the v2.41 arc added 10 ao subcommands; skills/validation/SKILL.md links references/flags.md with a real markdown link so heal.sh --strict passes; the cli-command-surface-matrix eval baseline + fixture updated to the current heading count (top=70 sub=173 all=243); and 7 /skillname slash-command references in skills-codex/{evolve,validation}/SKILL.md converted to Codex $skillname notation so the codex-native install canary passes.
  • --oscillation-sweep removal was incomplete in v2.41.0soc-1q1x deleted the ao defrag --oscillation-sweep flag but the cycle-181 audit only checked Go callers. Seven shell/YAML/doc callsites still referenced the removed flag, breaking the validate-flywheel-proof CI gate (scripts/nightly-dream-cycle.sh shelled the dead flag). Cleared from .github/workflows/nightly.yml, scripts/nightly-dream-cycle.sh, scripts/check-compile-oscillation.sh, docs/contracts/dream-run-contract.md, and the compile skill phase docs.
  • extract-release-notes.sh double-blank line — the awk extractor captured the blank line after the ## [VERSION] heading; wrapped in the <details> boilerplate it produced a double-blank between <summary> and the first subsection. Now strips leading/trailing blank lines from the changelog section.
  • ci-local-release.sh secret-scan false positive — the secret-pattern scan recursively grepped the gitignored _site/ MkDocs build directory, matching password:!0 inside minified JS bundles. Added _site and site to the scan's --exclude-dir list.

Full Changelog: v2.41.0...v2.41.1

v2.41.0

15 May 02:37

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v2.41.0 is the typed-BC-ports-load-bearing release. It completes the 14-port bounded-context inventory begun in the v2.40 cycle and makes the ports operationally load-bearing on the /evolve autonomous loop — not just scaffolding. Every /evolve cycle's read path now flows through typed Go interfaces (CorpusReaderPort Step 0, CIStatusPort Step 1.5, LoopReaderPort cycle-history wrapper) instead of untyped gh/ao lookup/inline jq shell-outs.

The release also ships 10 new ao subcommands that expose each typed port through the operator-facing CLI: ao loop history/append/verify, ao ci latest/recent, ao corpus inject/capture, ao operator record/list, ao harness status, ao gate run, ao citation verify, and ao claim bind/list. The cycle-shape used to land each of these is captured as a reusable template in docs/learnings/2026-05-13-cli-wiring-cycle-shape.md.

Three mechanical release-readiness gates now ship in the /validation, /evolve, and /post-mortem skills so the failure mode that produced this release — declaring "ready to tag" without running the full pre-push gate or regenerating CLI docs after a flag removal — cannot recur silently. STEP 1.7.5 of /validation auto-detects release context (branch name pattern or explicit flag) and refuses to recommend /release until scripts/pre-push-gate.sh (full, not --fast), scripts/ci-local-release.sh, and scripts/generate-cli-reference.sh cleanliness all pass. /evolve teardown on a release-shaped branch now emits an explicit pre-release checklist instead of recommending /release. /post-mortem closure-integrity now WARNs when a closed bead's acceptance text names a gate that the close-note doesn't confirm green.

A 448-line dead-code sweep removes defrag.SweepOscillatingGoals and its 5 helpers, the --oscillation-sweep CLI flag, and 9 corresponding tests (soc-1q1x path 1) — the function read for a target field that no production cycle has ever populated. Companion cleanup: 43 additional staticcheck U1000 findings cleared across cli/ over multiple cycles.

Upgrade Notes

  • CLI breaking-ish change: ao defrag --oscillation-sweep is removed. The flag was unused in production; downstream scripts that pass it will now error. If you depended on the report's Oscillation JSON field, it is also removed from DefragReport. The function never returned non-empty results in practice (see soc-1q1x analysis).
  • New typed-port CLI surface: 10 new ao subcommands are operator-facing and emit one JSON record per line. See cli/docs/COMMANDS.md for the full surface. The new commands route through typed BC port interfaces, so consumers get a stable shape independent of the on-disk file format.
  • Release-readiness gates are now mechanical: if you invoke /validation, /evolve, or /post-mortem on a release-shaped branch (release/*, v*-prep, v*-evolve-run, v[0-9]+.[0-9]+*), the new STEP 1.7.5 fires. Pass --skip-release-gates to bypass when running these skills on a release-shaped branch for non-release purposes.
  • The /release skill now requires the local CI parity gate (scripts/ci-local-release.sh) to pass before proceeding — --skip-checks does not skip this gate (changed in the v2.40.0 release cycle, hardened in v2.41).

At a Glance

Product Area Added Changed Fixed Removed
BC Ports (BC1–BC4) 4 0 0 0
Operator-facing CLI (ao) 11 1 0 1
Release-Readiness Gates (skills) 3 0 0 0
/evolve Read-Path Wire-up 1 3 0 0
Validation + Closure Integrity 1 0 1 0
Dead-Code Sweep 0 0 0 2
Internal: DDD/Hex Rescope Arc 0 1 0 0
Internal: Learnings Catalog 4 0 0 0

Product Areas

BC Ports — 14/14 Production-Complete

  • Added: FactoryAdmissionPort (13th BC port, soc-2klg.1) + productionFactoryAdmission pair adapter.
  • Added: ClaimEvidencePort (14th BC port, soc-2klg.2) + productionClaimEvidence.
  • Added: CycleEntry.StartedAt + Title fields (soc-ckc4) — ports widening + writer-side round-trip.
  • All 14 ports follow the <port>.go + inmemory_<port>.go + inmemory_<port>_test.go triplet with compile-time var _ XPort = (*InMemoryX)(nil) assertions as drift guards.

Operator-facing CLI (ao) Surface

  • Added: ao loop history (cycle 144 — first ao subcommand wired through a productionX adapter)
  • Added: ao ci latest/recent (cycle 145 — productionCIStatus)
  • Added: ao corpus inject (cycle 146 — productionCorpusReader; closes soc-y5vh.5)
  • Added: ao operator record/list (cycle 147 — template-applied 1st of 7)
  • Added: ao harness status (cycle 148)
  • Added: ao gate run (cycle 149)
  • Added: ao corpus capture (cycle 150 — BC1 R/W pair complete on CLI)
  • Added: ao loop append (cycle 151 — BC3 R/W pair complete on CLI)
  • Added: ao citation verify (cycle 152 — BC1 round-trip)
  • Added: ao claim bind/list (cycle 153 — 10th adapter CLI-exposed)
  • Added: ao loop verify — cycle-history integrity audit using cycle-161 widening
  • Removed: ao defrag --oscillation-sweep flag + DefragReport.Oscillation JSON field (soc-1q1x path 1)

Release-Readiness Gates (Skills)

  • Added: /validation STEP 1.7.5 — mandatory release-readiness gates with auto-detected release context, three gates (full pre-push, ci-local-release, generate-cli-reference cleanliness check), refusal to recommend /release until all pass. Codex parity synced.
  • Added: /evolve release-context teardown checklist — explicit unchecked pre-release checklist replaces the /release recommendation when the loop runs on a release-shaped branch.
  • Added: /post-mortem Acceptance-Text vs Delivered Drift audit — closure-integrity check that WARNs when a closed bead's acceptance language names a gate the close-note doesn't confirm green.

/evolve Read-Path Wire-up

  • Added: scripts/evolve-read-cycle-history.sh — wrapper around ao loop history for skill scripts.
  • Changed: Step 0 prior-knowledge retrieval routes through BC1 CorpusReaderPort via ao corpus inject (was: ao lookup).
  • Changed: Step 1.5 healing-first classifier routes through BC2 CIStatusPort via ao ci recent --limit 1 (was: inline gh run list).
  • Changed: oscillation counters in references/oscillation.md and references/fitness-scoring.md migrated to the wrapper (was: inline jq over raw JSONL).

Validation + Closure Integrity

  • Added: .agents/operator/ write-surface entry in docs/contracts/agents-write-surfaces.md (BC4 OperatorPort durable-intent log).
  • Added: Gap 3 (loop-closure) bats coverage in tests/scripts/check-three-gap-supergate.bats — 3 new tests (PASS / goals-validate FAIL / flywheel-proof SKIP). Suite: 15 → 18 tests.
  • Fixed: three-gap-supergate goals-validate sub-gate — pre-clean /tmp/ao-sg before go build -o /tmp/ao-sg so Go doesn't refuse to overwrite a non-object file. Caught by ao goals measure (1 failing → 0 failing).

Dead-Code Sweep

  • Removed: defrag.SweepOscillatingGoals + 5 helpers + CountAlternations + CycleRecord + OscillationResult + OscillatingGoal types + DefragReport.Oscillation field + defragOscillationSweep flag + 9 tests (soc-1q1x path 1, 17 / 465 ratio).
  • Removed: 43 additional staticcheck U1000 findings across cli/ (multiple cycles 156–159 sweeping soc-k083 backlog).

Internal: DDD/Hex Rescope Arc + Learnings Catalog

  • Changed: 13-cycle DDD/hex rescope arc closed; phase-1 complete + phase-2 retrospective (docs/rescope/2026-05-13-ddd-hex-architecture-rescope.md).
  • Added: BC-ports phase-2 narrowness post-mortem (docs/learnings/2026-05-13-bc-ports-narrowness-postmortem.md).
  • Added: empirical /loop context-drift study over 87+ cycles (docs/learnings/2026-05-13-loop-context-drift-87-cycle-observation.md).
  • Added: CLI-wiring cycle-shape template (docs/learnings/2026-05-13-cli-wiring-cycle-shape.md).
  • Added: learnings catalog README + when-to-add rubric.

See Also

  • Full per-commit detail: CHANGELOG.md (entries [2.41.0])
  • Release audit trail: docs/releases/2026-05-14-v2.41.0-audit.md
  • v2.40.0 notes: docs/releases/2026-05-13-v2.40.0-notes.md

Full changelog

Added

  • BC ports — 14/14 production-complete. FactoryAdmissionPort (13th, soc-2klg.1) + productionFactoryAdmission (cycle 139–140) and ClaimEvidencePort (14th, soc-2klg.2) + productionClaimEvidence (cycle 141–142) land the last two BC4 ports, closing the 14-port BC inventory begun in the v2.40.0 cycle. Each port follows the <port>.go + inmemory_<port>.go + inmemory_<port>_test.go triplet plus a productionX production adapter, with compile-time interface assertions as drift guards.
  • 10 typed-port-backed ao subcommands — operator-facing CLI now exposes every BC port that needs a script-side surface: ao loop history (productionLoopReader, cycle 144, slice 1 of soc-y5vh.5), ao ci latest/recent (productionCIStatus, cycle 145, slice 2), ao corpus inject (productionCorpusReader, cycle 146, slice 3; closes soc-y5vh.5), ao operator record/list (cycle 147 — template-applied 1st of 7), ao harness status (cycle 148), ao gate run (cycle 149), ao corpus capture (cycle 150, BC1 R/W pair complete on CLI), ao loop append (cycle 151, BC3 R/W pair complete on CLI), ao citation verify (cycle 152, BC1 round-trip), ao claim bind/list (cycle 153, 10th adapter CLI-exposed)....
Read more

v2.40.0

14 May 02:06

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v2.40.0 is the practice-citation + assurance-gates release. It
makes PRACTICE.md the derivation root for every primitive in the repo —
every skill, hook, eval suite, CLI command, and schema now declares its
practices: lineage, validated by an advisory CI gate. Twelve backfill passes
across skills/, hooks/, schemas/, evals/agentops-core/, and cli/
reached 756/756 declared primitives. AgentOps now produces a traceable
provenance graph from doctrine → primitive → evidence.

The release also lands the three-gap supergate (E5 epic), a unified
release-blocking gate consolidating contract floors, factory-admission
enforcement, and finding-registry coverage. Eight new CI-blocking contract
gates close the loop on factory-admission, finding-registry,
factory-yield-ledger, flywheel-compounding-snapshot, wiring-closure,
goals-validate, flywheel-proof, and quarantine-empty. The
contracts-structural-floor now covers all 38 contracts.

The behavioral eval workbench ships with 12 tasks, a live-agent suite, the
agent harness CLI, the context_comprehension dimension, a CDLC identity
field, and observability feedback. Three eval-related CI gates
(eval-skill-delta, eval-workbench-verify, head-to-head delta against the
baseline scorecard) make eval evidence part of the release surface.

This release adopts the three-layer product model (substrate / assurance /
bookkeeping) across the README, PRODUCT.md, the MkDocs landing page, and
positioning docs.

Upgrade Notes

  • Every new primitive added to skills/, hooks/, schemas/,
    evals/agentops-core/, or cli/ must now declare a practices: lineage.
    The CI gate runs as advisory initially; existing primitives are 100%
    declared via the backfill passes.
  • eval-suite manifests can declare a security domain. ao eval coverage
    will report security as missing until a security-domain manifest is
    authored — soft warning, not a release-blocker.
  • The five-minute first-value install gate (PG1) is now release-blocking. If
    install-e2e fails on Ubuntu or macOS, releases will block.
  • The corpus freshness gate (D11) is CI-blocking. Snapshot/restore is the
    recommended pattern for corpus recovery.
  • Three new commit-message warnings ship in hooks/: missing test-on-code,
    missing sibling-pattern citation, missing fitness-delta. All three are
    warnings, not blocks, in v2.40.
  • The /release skill was refactored under the 248-line size limit. If you
    customized the skill, re-apply your changes against the new
    references/release-workflow-detail.md extraction.

At a Glance

Product Area Added Changed Fixed
Practice-Citation + Assurance Provenance 16 0 1
Release Gates and CI 14 5 9
Eval Platform 9 0 6
Nightly Automation 8 0 1
Daemon and Factory 10 0 1
Skills and Workflows 6 2 2
CLI and Operator Commands 2 0 3
Hooks and Commit Discipline 3 0 2
Goals and Corpus 5 0 1
Docs, Positioning, and Product Framing 4 8 1
Dependencies 0 12 0

Product Areas

Practice-Citation + Assurance Provenance

  • Added: practice-citation derivation graph + advisory CI gate
    (254138fa) — every primitive now declares its practices: lineage.
  • Added: 12 backfill passes (pass-1 through pass-12, with four pass-12
    waves across cli/) reached 756/756 declared across skills/,
    hooks/, schemas/, evals/agentops-core/, and cli/.
  • Added: 10 chore(rpi) consumption commits drove the backfill epic
    end-to-end.
  • Added: corpus-stats.sh + derived PRODUCT.md evidence (epic
    soc-sx99 W3.3).
  • Fixed: practice-provenance validator gaps (9e981d59).

Release Gates and CI

  • Added: three-gap supergate (E5 epic, d8815e5f) with --strict-coverage
    opt-in (335a70a6).
  • Added: contracts-structural-floor for all 38 contracts (f822aa38).
  • Added: factory-admission contract gate (355313cc), finding-registry
    contract gate (ed6cacbf), factory-yield-ledger contract gate
    (c0b6d09d), flywheel-compounding snapshot gate (fafa00e7),
    wiring-closure gate (b56f4def), goals-validate gate (57ca74a7),
    flywheel-proof gate (491ff96b), quarantine-empty gate (55a85314),
    contract canaries gate (42f57e36), check-docs-learning-references
    gate (9623d6b6).
  • Added: install-e2e workflow against ubuntu+macos (0c0b79e2).
  • Added: codex-parity-drift CI gate (09b44d79).
  • Added: path-filter conditions to remaining 28 CI jobs (ce4c0150).
  • Added: inner/middle/outer loop boundary enforcement (1e0317f8, #230).
  • Added: bats path-filter wiring (3c8b33ae); bats prints captured output
    on failure (e6f7586e).
  • Changed: --two-pass mode added to gate runner (5dbe1e6d); local scope
    default to head.
  • Fixed: registry non-determinism + pre-push goals-validate fallback;
    structural CI failures from cycles 45-47; CLI surface count drift after
    patterns/repair-filenames addition; registry.json knowledge_stores +
    schedules must match gitignored-state regen.

Eval Platform

  • Added: behavioral eval workbench with 12 tasks and first suite
    (72b181b8); workbench expanded to all 12 (4ef6cf5e).
  • Added: live agent eval suite with 3 workbench cases (b323af2d).
  • Added: agent harness script wired into the eval CLI (dbbe159d);
    industry-proven eval patterns added (e5defbb8).
  • Added: context_comprehension dimension + CDLC identity + observability
    feedback loop (80f53fb0).
  • Added: head-to-head delta gate against baseline scorecard (D10,
    8dcfd92d).
  • Added: eval-skill-delta CI gate + nightly schedule template (9ba08146).
  • Added: eval-workbench-verify gate added to GOALS.md (Directive 10,
    32840177); eval-workbench-verify CI job (7c0ab19b).
  • Added: eval run records uploaded as CI artifacts for triage (c3eefb0d);
    Python venv bootstrap in agentops-eval-advisory (9f878a2f).
  • Added: security domain in eval-suite manifests (schema + production
    default + test fixture in lock-step).
  • Fixed: six fix(eval) patches covering canary refresh, advisory job
    stability, and workbench drift.

Nightly Automation

  • Added: admission-aware morning digest with durability (4e0dbecf).
  • Added: manual-PR-only landing policy for evolve (52ff0c4c).
  • Added: fail-closed execute preflight for source mutation (7384e33c).
  • Added: blocker matrix + main CI baseline artifacts (3e4aa755).
  • Added: dream runs submitted through the daemon (a2bea16f).
  • Added: L3 rehearsal scenario + operator runbook (e47be280).
  • Added: scheduler install helper + updated runbook (b10b4354).
  • Added: PR digest generator from structured run state (54bd161a).
  • Fixed: dream-degraded added to next-work source enum (f02f0856).

Daemon and Factory

  • Added: factory-admission job specs (8281ecfd), executor (3f907375),
    routing-lane validators (634ff343).
  • Added: daemon RPI agent-update events on phase boundaries
    (soc-y0ct.2, 94056823, #258) and agent_update.criterion_verdict
    per wave checkpoint (soc-awx8, 72abc18a, #259).
  • Added: daemon skill schedule wiring (28da44fc).
  • Added: CLI fallback RPI executor (45dd2825).
  • Added: eval and planning job-type registration (1022c741).
  • Added: factory claim ledger with proof taxonomy.
  • Added: Wave 1A-D factory-claim-ledger reconciliation (f331c036,
    soc-e4ulx, #264).
  • Added: RPI lifecycle sharpening — criterion contract + isolation
    enforcement + daemon executor swap (729fbf8a, soc-bcrn, #255).
  • Fixed: bound daemon RPI GasCity sessions.

Skills and Workflows

  • Added: skill-builder + skill-auditor pair (epic soc-9bak, #237).
  • Added: ao quickstart surfaces ao schedule and ao daemon
    (epic soc-sx99 W2.2, e96d4942).
  • Added: tracer-bullet shape for skills/domain ubiquitous-language
    corpus (28fd87ad).
  • Added: scoped evidence on /crank and /implement bead closures
    (1c783e70).
  • Added: 12 standalone skills consolidated into beads, review, doc,
    research.
  • Added: JSM quality playbook slices absorbed in two waves (268e43b5,
    b6e03c71).
  • Changed: /release skill refactored under the size limit + non-HEAD
    cut support (5cbab078) — the commit that prepared the v2.40 cut.
  • Changed: skill counts synced to 71 after consolidation.
  • Fixed: unblock CI for domain skill — registry, codex manifest, catalog
    budget (b75da352).

CLI and Operator Commands

  • Added: ao session spawn — template-driven session launch
    (b27e527f, #234 via feat/session-spawn).
  • Added: ao feedback-loop --drain — clear unfed citation backlog
    (c84e190c, epic soc-sx99 W3.1).
  • Fixed: three fix(cli/rpi) correctness fixes around RPI scoping and
    isolation enforcement.

Hooks and Commit Discipline

  • Added: hook warns on commits with code-without-test (P2, 9e9de5e9).
  • Added: hook warns on commits missing sibling-pattern citation (P3,
    a16add91).
  • Added: hook warns on commit messages missing fitness-delta (P4,
    ecb3b3ba).
  • Fixed: stale .agents/.gitignore deny-all reconciled with parent
    allowlist (soc-rv5p, #263); pre-push refs-transmission proof (#254).

Goals and Corpus

  • Added: corpus snapshot/restore + freshness gate (D11, 7358ec62).
  • Added: goals summary split into code-driven vs runtime-artifact
    (1ab6bbb2).
  • Added: goals SKIP exit code 77 + flywheel-compounding dormant
    precondition (1ed54715).
  • Added: AffectsFiles sidecar wired for the ...
Read more

v2.39.0

04 May 04:03

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v2.39.0 is the daemon and CLI architecture release. It introduces
agentopsd as an opt-in local control plane for durable jobs, scheduling,
worker execution, projections, and operator status, while keeping the existing
foreground RPI, Dream, wiki/forge, and plans flows available.

This release also reorganizes the Go CLI around clearer internal subsystems and
expands the command surface for daemon jobs, schedule management, evals,
agent-state health, skill health, edit scopes, pool maintenance, and
daemon-backed workflows.

Upgrade Notes

  • Daemon mode is opt-in. Existing foreground commands continue to work unless
    you pass the new daemon flags or run ao daemon.
  • If you use the Codex native plugin, refresh the plugin and restart Codex to
    pick up the updated metadata, hooks, and $skill invocation shape.
  • Repo-root .agents/ runtime state is no longer tracked by git. Keep durable
    release artifacts and docs outside .agents/.

At a Glance

Product Area Added Changed Refactored Fixed Deprecated/Removed
Install, Upgrade, and Distribution 1 1 0 1 1
CLI and Operator Commands 8 4 3 5 0
Daemon, Scheduling, and Factory 9 5 3 9 0
Skills and Workflows 5 4 1 4 0
Codex and Runtime Integrations 2 4 1 5 1
Hooks and Lifecycle 3 3 2 5 0
Knowledge Flywheel, Search, and Memory 5 4 2 4 0
Eval, Validation, and Release Gates 8 7 2 7 0
Docs and Onboarding 6 5 0 4 0
Security, Privacy, and Supply Chain 3 2 0 4 0
Contributor/Internal Refactors 4 3 6 6 0

Product Areas

Install, Upgrade, and Distribution

  • Added: scripts/install.sh --dev for one-command development bootstrap.
  • Changed: Codex native plugin metadata now matches the marketplace schema and
    ships the refreshed checked-in Codex skill bundle.
  • Fixed: release validation no longer trips over stale audit metadata or
    generated Codex artifact drift during publish prep.
  • Deprecated/Removed: the old tracked repo-root .agents/ state was removed
    from git history and replaced with local runtime-state guidance.

CLI and Operator Commands

  • Added: ao daemon commands for readiness, status, event tailing, job
    submission, job inspection, cancellation, waiting, service-plan output, and
    product soak scenarios.
  • Added: ao schedule commands for adding, listing, running, and removing
    recurring daemon jobs.
  • Added: ao agents inspect, ao agents lint, and ao agents doctor for
    .agents/ write-surface introspection and health checks.
  • Added: ao skills check, ao scope, ao pool reindex, ao watch, and new
    ao eval task/cleanup/suite/baseline surfaces.
  • Changed: RPI, Dream, plans, and wiki/forge commands gained explicit daemon
    submission or daemon-read paths while preserving foreground fallbacks.
  • Refactored: large parts of ao moved out of command glue into focused
    internal packages for daemon state, scheduling, workers, GasCity, eval,
    lifecycle, path resolution, harvest, and knowledge flows.
  • Fixed: JSON output, UTF-8 truncation, Cobra command docs, hidden command
    coverage, and command/test pairing were tightened across the CLI.

Daemon, Scheduling, and Factory

  • Added: agentopsd, an opt-in local daemon with a durable ledger, local HTTP
    API, job queue, worker supervisor, event stream, readiness/status endpoints,
    mutation tokens, and rebuildable projections.
  • Added: daemon job families for rpi.run, rpi.phase, dream.run,
    dream.stage, wiki.forge, wiki.build, openclaw.snapshot, and
    plans.projection.
  • Added: .agents/schedule.yaml, cron parsing, backpressure, schedule mutation
    routes, recurrence supervisor logic, and schedule ledger events.
  • Added: AgentWorker, GasCity, CLI-fallback worker, cgroup, routing-policy,
    factory lifecycle, worktree ownership, validation, manual-merge, and yield
    contracts.
  • Changed: daemon-accepted work now treats .agents/daemon/ledger.jsonl as the
    source of truth and projections as rebuildable read models.
  • Refactored: daemon mutation handling, projection rebuilds, recurrence,
    RPI/Dream/wiki executors, and factory status projection were split into
    smaller state-machine pieces.
  • Fixed: fsync boundaries, orphan temp sweeps, snapshot sync, idempotency
    dedupe, queue cancellation, recurrence recompute, projection nil-safety,
    claimed-job queue depth, terminal-state precedence, and worker cancellation.
  • Fixed: daemon API hardening for body caps, event limits, bad cursors,
    malformed schedules, oversized ledger lines, path containment, and schedule
    name traversal.

Skills and Workflows

  • Added: system-tuning and scope skills, plus branch-isolation and
    parallel-wave references for larger crank runs.
  • Added: council presets and symmetric mixed-vendor pairing behavior for
    multi-model reviews.
  • Changed: release, validation, vibe, RPI, and using-agentops skills were
    updated for the new Codex runtime and validation surfaces.
  • Changed: bootstrap now recommends installing bd instead of attempting an
    automatic install.
  • Refactored: skill reference structure and tier-cap guidance were tightened so
    large skills can stay under maintainable size limits.
  • Fixed: skill integrity, quickstart aliases, Codex twins, language-gated
    complexity checks, and several smoke prompt expectations.

Codex and Runtime Integrations

  • Added: refreshed Codex native plugin metadata and checked-in Codex artifacts
    for GPT-5.5-era runtime behavior.
  • Added: headless runtime and native Codex install validation coverage for the
    plugin, hooks, runtime sections, backbone prompts, and override coverage.
  • Changed: Codex skill references now use $skill notation and a smaller skill
    catalog context footprint.
  • Changed: native Codex hooks install into the plugin cache with the expected
    22 handlers across 5 supported events.
  • Refactored: Codex hook/runtime proof paths moved toward explicit native
    plugin validation instead of legacy raw skill mirrors.
  • Fixed: Codex skill chaining defaults, native hook manifest installation,
    quiet session-start behavior, hook test gaps, lifecycle guards, artifact
    metadata, and stale hook-cache diagnostics.
  • Deprecated/Removed: routine ao codex start/stop lifecycle shims are now
    documented as deprecated for normal validation and merge work.

Hooks and Lifecycle

  • Added: hook output schema linting for Claude/Codex portable PreToolUse output.
  • Added: edit-scope guard, edit/write hash-audit hooks, and verdict compiler
    hook support.
  • Changed: hooks now use managed/JIT runtime context and shared path helpers
    instead of scattering repo-state path logic.
  • Refactored: hook helper layout now sources lib/ao-paths.sh, and embedded
    hook copies were resynchronized.
  • Fixed: noisy Codex session-start output, hook manifest drift, intent-echo
    bypass for team runners, pre-push hook coverage, and hook/docs parity.

Knowledge Flywheel, Search, and Memory

  • Added: .agents/ write-surface contract, lint gate, operator guide, and
    production-code smoke coverage for allowlisted runtime surfaces.
  • Added: Dream finding-generator sidecars, aggregation, external-watchlist
    output, and end-user coverage fitness gates.
  • Added: native harvest extraction, nested artifact traversal, real rig
    metadata, and pool reindexing.
  • Changed: RPI next-work metadata now treats status, requires, dedup_key,
    and external-watchlist routing as first-class fields.
  • Changed: goals measurement can exclude long-cycle corpus-state tags when a
    gate should ignore dormant long-cycle state.
  • Refactored: flywheel and pool ingestion paths were split into shared helpers
    with stronger per-file and promoted-body dedupe tests.
  • Fixed: harvest TOCTOU handling, path walk-up behavior, stale next-work
    bookkeeping, dry-run packet aliases, close-loop replay stability, and
    triple-ID amplification in close-loop pending IDs.

Eval, Validation, and Release Gates

  • Added: deterministic ao eval core, runtime adapters, scorecards, coverage
    reporting, baseline A/B, context-packet A/B, task/cleanup substrate commands,
    retrieval/file-backed backends, and public canary suites.
  • Added: contract canaries as a required gate and modular Codex runtime jobs in
    CI.
  • Added: release readiness, audit-artifact validation, HIL/SIL/VIL evidence,
    test-fixture parity, retrieval manifest validation, and pre-push-gate wiring
    checks.
  • Changed: baseline-audit is now drift-only for promoted suite hashes, with
    policy mismatch reported separately.
  • Changed: nightly knowledge-cycle skips dormant corpora as one precondition
    instead of failing multiple downstream jobs.
  • Changed: advisory CI policy, Windows smoke, eval advisory setup, and
    retrieval-bench promotion rules are now documented and enforced more
    consistently.
  • Refactored: eval cleanup, task run, validation, and gate parsing logic were
    split into lower-complexity helpers.
  • Fixed: eval fixtures, canary counts, baseline path handling, coverage anchor
    parsing, shellcheck regressions, Windows path handling, and release gate drift.

Docs and Onboarding

  • Added: daemon migration, daemon scheduling, control-plane, routing-policy,
    local compute, OpenClaw, GasCity, cloud frontier, and factory runbooks.
  • Added: .agents/ operator guidance, competitive comparisons, release audit
    references, and component-level code maps for daemon/eval sur...
Read more

v2.38.0

23 Apr 00:11

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

AgentOps v2.38.0 hardens the orchestrator skills against a "compression"
anti-pattern that was silently inlining phase work, rebuilds the
documentation site on MkDocs Material with generated skill and CLI
references, and unblocks the overnight flywheel's promotion pipeline
that had been quietly stuck on a citation-gate deadlock. Ergonomics
improvements include shell tab-completion for enumerated flags and a
faster ASCII-input path for the CLI's truncation helpers.

What's New

  • Orchestrator skills resist compression/rpi, /discovery,
    /validation, and /crank now declare strict sub-skill delegation as
    the default, with documented escape flags (--quick, --fast-path,
    --no-retro, etc.) for when you deliberately want to skip a phase.
  • New documentation siteagentops.dev-style site rebuilt on
    MkDocs Material with a dark terracotta palette. The skill catalog and
    CLI reference are now generated at build time, so they can't drift
    from the source. New pages cover hooks, schemas, and upgrading.
  • Close-loop promotion unblocked — an overnight citation-gate
    deadlock had been silently zeroing throughput for weeks. The gate now
    fails loudly when ingests happen without promotions, and the
    underlying cycle is fixed.
  • Shell tab-completionao <cmd> --<flag> <TAB> now suggests the
    valid values (e.g. json/table/yaml) instead of listing files.
  • Faster truncation on ASCII input — the CLI's text-truncation
    helpers skip the UTF-8 rune scan when the input is plain ASCII.

All Changes

Added

  • Strict delegation contract across /rpi, /discovery, /validation,
    /crank with a shared reference document
  • Orchestrator compression anti-pattern learning surfaced at session
    start
  • MkDocs Material docs site with generated skill catalog and CLI
    reference; new Hooks / Schemas / Upgrading pages and expanded glossary
  • Shell completion for enumerated-value flags
  • ao doctor now scans Codex skill mirrors and skill reference docs for
    stale commands
  • Nightly throughput alarm that fails loudly on dream close-loop stalls

Removed

  • Archived cross-system bridge integration that predates the current
    runtime
  • Duplicate lowercase docs landing page (canonicalized during the
    MkDocs migration)

Changed

  • /discovery flag renamed: --no-lifecycle is now --no-scaffold
    (old name still honored through v2.40.0)
  • ASCII fast-path performance sweep across CLI truncation call sites
  • Compile and overnight internals refactored for lower complexity and
    tighter test seams (no behavior change)
  • GitHub Actions bumped (pages artifact, pages deploy, pages configure)
  • Precommit hook prefers local Go when available

Fixed

  • Overnight findings router now emits schema-valid work-queue enums and
    has a build-time guard against future drift
  • Stale-refs scan no longer false-positives on rename notes
  • Release gates self-heal missing complexity checker; compile gates
    fall back to a preview path when the primary is unavailable
  • Proof runs now ingest before closing the loop
  • Hook and script edges — worker git guard scope, test-hook tolerance
    for missing binaries, scripts that skipped cleanly after the
    goals-file migration
  • CI recovery from the docs rebuild (7 failures resolved)
  • Docs markdownlint violation in a harvest-plan doc

Full changelog


Full changelog

Added

  • Strict Delegation Contract for /rpi, /discovery, and /validation — top-level orchestrator skills now declare strict sub-skill delegation as the default. Each skill points to the new canonical reference skills/shared/references/strict-delegation-contract.md which documents the contract, anti-pattern rationalizations, and supported compression escapes (--quick, --fast-path, --no-retro, --no-forge, --skip-brainstorm, --no-scaffold, --no-behavioral, --allow-critical-deps). There is no --full flag — strict delegation is always on.
  • Orchestrator Compression Anti-Pattern learning at .agents/learnings/2026-04-19-orchestrator-compression-anti-pattern.md, surfaced via ao inject at session start. Includes detection phrases, corrective actions, and rationalizations to reject.
  • Orchestrator-owned step markers in skills/crank/SKILL.md (STEP 3a.3, STEP 6.5 slop-scan, STEP 8.7) plus an "Inline Work Policy" footer documenting which steps are intentionally inline vs delegated.
  • MkDocs Material documentation site — Pages site rebuilt on MkDocs Material (slate dark palette). Skill catalog and CLI reference are generated at build from skills/*/SKILL.md and cli/docs/. mkdocs build --strict wired into the pre-push gate (Check 25a). New dedicated pages for hooks, schemas, and upgrading, plus an expanded glossary. Theme tuned to the agentops-showcase terracotta-on-near-black palette; landing page leads with the primary use case and a headline skills table; flywheel diagram ASCII art realigned; doctrine back-links added to 12factoragentops.com.
  • Shell completion for enumerated-value flagsao <cmd> --<flag> <TAB> now suggests the valid values instead of falling back to file completion. Covers ao --output (json/table/yaml), ao seed --template, ao goals init --template, ao inject --format and --session-type.
  • ao doctor stale-reference scan now covers skills-codex/*/SKILL.md and skills/*/references/*.md in addition to the primary skill docs, catching drift in Codex mirrors and skill reference content.
  • Nightly close-loop throughput alarm — the dream close-loop gate now fails loudly when ingested > 0 && promoted == 0, replacing the silent zero-throughput mode that previously masked citation-gate deadlocks.

Removed

  • Archived AO↔Olympus bridge integration: removed docs/ol-bridge-contracts.md, docs/architecture/ao-olympus-ownership-matrix.md, MemRL policy contracts, skills/*/scripts/ol-*.sh, cli/cmd/ao/inject_ol_test.go, and associated CLI types (OLConstraint, gatherOLConstraints, .ol/ directory collector). Olympus predecessor's useful patterns live on inside ao.
  • Lowercase docs/index.md duplicate removed after the MkDocs migration canonicalized the landing page.

Changed

  • --no-lifecycle in /discovery renamed to --no-scaffold for semantic clarity — the flag controls STEP 4.5 scaffold auto-invocation only, not broader lifecycle checks. --no-lifecycle is honored as a deprecated alias through v2.40.0; when both flags are passed, they are equivalent. Other skills (/crank, /validation, /implement, /evolve) retain --no-lifecycle with its existing lifecycle-skill-invocation semantics.
  • /discovery flags table expanded: --auto is now explicitly documented (was transitively honored but undocumented); --interactive scope clarified ("research + plan gates, not pre-mortem").
  • /validation flags table expanded: --complexity=<level> syntax formalized to match /rpi and /discovery; --interactive scope documented.
  • /rpi --interactive flag scope note added: applies to discovery (research + plan) and validation (Gate 1, Gate 2); does NOT override pre-mortem or vibe council autonomy.
  • ASCII fast-path performance sweep across rune-aware truncation call sites in cli/ (TruncateText, TruncateRunes, truncateForError, plus goals/pool/search/rpi/parser call sites) — ASCII inputs now skip the full UTF-8 rune scan.
  • Compile and overnight internals refactoredrunCompile split into phase + preflight helpers; article scan, inbound count, and prune extracted from repair; dream packet corroboration split per source epic; dream yield emptiness guard extracted into a dedicated helper. No behavior change; lower cyclomatic complexity and tighter test surfaces.
  • Skills-codex DAG bodies converted to $skill notation for the Codex runtime.
  • GitHub Actions bumpedactions/upload-pages-artifact 3→4, actions/deploy-pages 4→5, actions/configure-pages 5→6, plus docs.yml workflow action versions aligned.
  • Precommit hook prefers local Go when available.
  • Skills backfill pass — docs, validators, and lint synced across all skills; Codex drift surface reset to no-op.

Fixed

  • Orchestrator compression vulnerability — a live compression was observed 2026-04-19 where /rpi was invoked but phases were inlined instead of delegated. This release documents the anti-pattern (forged learning + loud skill text), scaffolds future enforcement (shared contract reference used by all 6 orchestrator skills), and explicitly defers runtime hook enforcement to a follow-up initiative. It does not mechanically prevent compression yet — the durable fix depends on ao inject surfacing the forged learning on future session starts. See .agents/research/2026-04-19-rpi-skill-dag-audit.md for the audit and .agents/plans/2026-04-19-rpi-dag-hardening.md for the remediation plan.
  • Close-loop promotion deadlockflywheel close-loop auto-promotion and the loop-dominance signal are unblocked; citation-gate cycles no longer silently zero throughput.
  • Overnight findings router now emits schema-compliant next-work v1.3 enums — valid claim_status=available (was pending), severity collapsed to high for critical and blocker inputs — with a build-time guard that fails on future drift.
  • Quality stale-refs scan skips rename-doc lines so it no longer false-positives on deliberate rename notes.
  • Release and compile gates — `go-comp...
Read more

v2.37.2

16 Apr 00:22

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

This hotfix hardens AgentOps' validation and execution surfaces across hooks,
Codex runtime artifacts, and release gating. It also adds proof-backed swarm
evidence checks and safer worker git defaults, while tightening several CLI,
compile, and harvest edges that surfaced during release prep.

What's New

  • Proof-backed swarm validation — release and validation gates now treat
    swarm evidence as a first-class contract.
  • Safer multi-agent git behavior — worker sessions now respect a
    lead-only git guard in the hook chain.
  • More recoverable compile and harvest runs — compile adds reset and
    repair controls, and harvest explains near-miss exclusions instead of
    silently dropping them.

All Changes

Added

  • Swarm-evidence schema and validator, wired into validation and release gates
  • Lead-only worker git guard in the hook chain for multi-agent sessions
  • Compile runtime preference plus --reset and --repair, and harvest
    near-miss reporting

Changed

  • Release, pre-push, and Codex/runtime validation now cover more hook,
    evidence, and artifact surfaces before publish
  • Next-work backlog bookkeeping and Codex/runtime docs were normalized to
    better match shipped behavior

Fixed

  • Pre-mortem gate ambiguity now fails closed instead of open
  • ao rpi serve --run-id, ao mine --dry-run, compile and harvest edge
    handling, CI fixture drift, shellcheck drift, and Codex artifact metadata
    drift

Full changelog


Full changelog

Added

  • Swarm evidence validation — AgentOps now ships a swarm-evidence schema and validator, and wires that proof surface into validation and release gates.
  • Lead-only worker git guard — worker sessions now have an explicit lead-only git guard in the hook chain, reducing accidental write authority in multi-agent runs.
  • Compile and harvest operator controlsao compile adds runtime preference plus --reset and --repair controls, while harvest now reports excluded low-confidence candidates and top near-misses.

Changed

  • Release and pre-push validation — local release, pre-push, and command coverage gates now validate more of the hook, evidence, and Codex runtime surface before publish.
  • Codex/runtime artifacts and docs — compile, evolve, post-mortem, swarm, and related runtime docs and artifacts were decomposed and synchronized to better match shipped behavior.
  • Flywheel backlog bookkeeping — next-work aggregates, consumed markers, and enum normalization were cleaned up so carry-forward work is recorded consistently.

Fixed

  • Pre-mortem gate ambiguity — the crank pre-mortem gate now denies ambiguous state by default instead of failing open.
  • CLI and shell reliability edgesao rpi serve --run-id now accepts legacy 8-hex IDs, ao mine --dry-run emits a single clean JSON payload, and bash invocations are sanitized to bypass unsafe shell aliases.
  • Compile, harvest, and release drift — compile repair defaults, malformed frontmatter salvage, YAML parse error surfacing, CI fixture drift, shellcheck drift, and Codex artifact metadata drift were corrected.

Full Changelog: v2.37.1...v2.37.2

v2.37.1

15 Apr 04:24

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

Dream now leaves behind actionable morning work instead of just a short
overnight summary. This hotfix adds ranked morning packets, an opt-in long-haul
corroboration pass for weak results, and a reliable headless Claude path for
Dream Council, so the overnight loop is both more useful and more dependable.

What's New

  • Actionable morning packets — Dream can turn overnight evidence into
    ranked next steps with suggested commands and handoff metadata.
  • Smarter long-haul mode — extra overnight time is only spent when the
    first pass is weak, and it is used to corroborate the strongest packet.
  • Reliable headless Claude council — Dream Council now works against the
    real Claude JSON contract instead of the broken spawn path that degraded or
    timed out runs.

All Changes

Added

  • Dream morning packets that carry ranked next steps, evidence, target files,
    and queue or bead handoff metadata
  • Yield telemetry and an adaptive long-haul mode that can corroborate weak
    overnight output before handing it off

Changed

  • Overnight runs now prefer cheaper evidence corroboration before slower
    council fan-out, so strong runs stay short and only weak output pays the
    extra runtime cost

Fixed

  • Headless Claude council spawning and result normalization for Dream
  • Overnight close-loop reporting so Dream writes real report artifacts instead
    of placeholder statuses
  • Retrieval-quality release gating so local release checks can fall back to
    checked-in eval data when a local manifest is missing

Full changelog


Full changelog

Added

  • Dream morning packets — Dream can now emit ranked morning work packets with evidence, target files, exact follow-up commands, and queue/bead handoff metadata.
  • Dream yield telemetry and long-haul corroboration — overnight reports now record packet-confidence telemetry and can trigger a bounded long-haul corroboration pass when the first pass produces weak morning output.

Changed

  • Dream decision flow — overnight runs now prefer cheaper evidence corroboration before slower council fan-out, so strong runs stay short and extended runtime is reserved for genuinely weak output.

Fixed

  • Headless Claude Dream council — Dream now uses Claude's working JSON output contract for headless council runs and normalizes the returned envelope before validation.
  • Dream close-loop and report surfaces — overnight runs now write real close-loop callbacks and post-loop report artifacts instead of leaving placeholder pending steps.
  • Retrieval ratchet release gate fallback — the retrieval-quality release check now falls back to checked-in eval data when a local manifest is absent.

Full Changelog: v2.37.0...v2.37.1

v2.37.0

14 Apr 22:11

Choose a tag to compare

brew update && brew upgrade agentops · bash <(curl -fsSL https://raw.githubusercontent.com/boshu2/agentops/main/scripts/install.sh) · checksums · verify provenance


Highlights

This release pushes AgentOps further toward a repo-native knowledge workspace.
With Karpathy's LLM-wiki getting attention, it's clearer that AgentOps was
already converging on a similar shape. This release takes some of the pieces he
executed better, especially index-first wiki navigation and knowledge
ergonomics, and integrates them into the existing .agents workflow.

At the same time, AgentOps now has a real Windows install path, a first-class
ao compile command, and a local LLM pipeline that can turn session
transcripts into reviewable wiki pages. Dream and evolve also connect more
tightly, so overnight knowledge passes can feed the daytime improvement loop
instead of living beside it.

What's New

  • Windows is a real target — PowerShell install plus blocking native
    Windows smoke coverage.
  • Knowledge compilation is a CLI commandao compile moves the compiler
    out of skill-only territory.
  • Transcript-to-wiki pipelineao forge can redact, summarize, review,
    and store local LLM session pages in a browsable .agents wiki.
  • Dream feeds evolveao evolve --dream-first and --dream-only let
    knowledge passes run before or instead of code cycles.
  • Stronger planning and retrieval guardrails — beads audit and cluster
    surfaces, retrieval evaluation fixtures, and a CI ratchet make stale or weak
    retrieval behavior easier to catch.

All Changes

Added

  • Windows installer and native Windows smoke validation for install, doctor,
    and overnight-sensitive paths
  • ao compile command with docs and tests
  • Local LLM forge pipeline with redaction, structural review, and Dream
    integration
  • .agents wiki scaffolding with INDEX, LOG, and searchable wiki directories
  • New beads audit and clustering helpers, status quality signals, and retrieval
    quality ratchets

Changed

  • Dream can now run as a knowledge-first sub-cycle inside evolve
  • Search and injection rank knowledge more intelligently using content dedup,
    index boosts, and stability weighting
  • Public docs and workflow guidance now line up better with the current
    operational-layer and context-compiler story

Fixed

  • Windows overnight liveness detection
  • Release retag safety and audit artifact validation
  • Post-mortem closure audit handling for evidence-only and path-heavy cases
  • Several Codex and RPI reliability edge cases around lifecycle restarts, JSON
    writes, proof paths, and bridge validation

Full changelog


Full changelog

Added

  • Windows install and smoke coveragescripts/install-ao.ps1 adds a first-class Windows install path, and the blocking windows-smoke gate exercises PowerShell install, local ao doctor, and Windows-sensitive Go packages.
  • Compile commandao compile makes knowledge compilation a first-class CLI surface with docs and tests.
  • Local LLM forge pipelineao forge can now redact, summarize, structurally review, and queue transcript-derived wiki pages with Dream worker integration.
  • Dream curator and evolve sub-cycle — Dream gained a local curator adapter plus ao evolve --dream-first|--dream-only, allowing overnight knowledge passes to feed the daytime improvement loop.
  • .agents wiki surfaces — INDEX, LOG, wiki directories, and search integration formalize .agents/ as a Karpathy-style knowledge wiki with index-first navigation.
  • Operational quality surfaces — beads audit/cluster commands, swarm preflight advice, status quality signals, retrieval eval queries, and a retrieval-quality CI ratchet broaden release-time proof.

Changed

  • Knowledge scoring and search behavior — inject now deduplicates by content hash, boosts indexed pages, weights stability, and search can pull Dream vault and wiki sources with stronger local recall.
  • Overnight and RPI internals — overnight, lifecycle, search, inject, harvest, and RPI flows were decomposed into smaller helpers while tightening proof paths, mixed-mode provenance, and worktree cleanup.
  • Public framing and contributor docs — README, philosophy, planning/post-mortem docs, and reference surfaces now better match the context-compiler and operational-layer story.

Fixed

  • Windows overnight liveness — Windows process checks no longer rely on Unix signal(0) semantics.
  • Dream RunLoop status invariants — live-tree hash coverage now exercises every terminal RunLoop status, and degraded reflects the current rollback semantics.
  • Release retag safety — release tooling now preserves annotated tags, validates audit artifact manifests and refs, and cancels stale reruns before duplicate publish attempts.
  • Post-mortem and closure audits — metadata links, evidence-only closure packets, parser-path handling, and closure packet evidence modes were normalized.
  • Codex and runtime reliability — same-thread lifecycle restart, root-scoped fallback reads, JSON config writes, bridge contract validation, and next-work proof-path handling were hardened.

Full Changelog: v2.36.0...v2.37.0